brief communications Arsenic poisoning in the delta The natural contamination of drinking water by arsenic needs to be urgently addressed.

he pollution by naturally occurring 0 arsenic of alluvial Ganges aquifers, Twhich are used for the public water supply in and West , has 20 been discussed by Nickson et al.1. We agree with their main conclusion that arsenic is 40 released by reductive dissolution of iron oxyhydroxides, as was proposed earlier2. Our observations indicate that arsenic- 60

rich pyrite and other arsenic minerals, Depth (m) which were proposed in previous models 80 (cited by Nickson et al.1) to give rise to arsenic pollution, are rare or even absent in 100 the sediments of the Ganges delta. We believe that arsenic is more likely to be co- precipitated with or scavenged by iron (III) 120 0 100 200 300 400 500 600 700 800 900 and manganese (IV) in the sedimentary Figure 1 Map showing arsenic-affected areas (shaded). Cu, cop- Total arsenic (µg per litre) environment. per belt of Bihar; S and D, Subarnarekha and Nickson et al.1 suggested that the arsenic basins; C, coal basins; R, Rajmahal volcanics; H, ; and Figure 2 Variation of total arsenic with depth in aquifers in in these alluvial sediments is derived from SO, sulphide occurrence with arsenic in Himalayas. . sulphide deposits in the . However, the copper belt of Bihar, which The relatively low values of dissolved iron increased the natural mobility of arsenic. contains small amounts of arsenopyrite, upstream of the Ganges delta indicate that The efficiency with which arsenic is and the coal basins of the Damodar valley, the environment may not be sufficiently removed by adsorption onto iron-coated which contain moderate concentrations of reducing to mobilize iron and arsenic. sand10 and by adsorption on and co-precip- arsenic, are drained by rivers that flow far to Nickson et al.1 reported that arsenic con- itation with ferrihydrite11 depends on both the south of the Ganges tributary system centration increases with depth in wells at the arsenic oxidation state and the ratio of (Fig. 1). We suggest that there are several Manikganj, Faridpur and Tungipara in iron to arsenic. The proposed removal of more likely sources of sedimentary arsenic, Bangladesh. However, this observation arsenic by simple aeration1 of anoxic water including the Gondwana coal seams in the appears to be site specific, as the large data- must therefore be approached with caution. Rajmahal basin, which contain up to 200 base5–7 for aquifers in West Bengal indicates S. K. Acharyya*, P. Chakraborty*, parts per million (p.p.m.) of arsenic; isolated that arsenic decreases with depth (Fig. 2). S. Lahiri*, B. C. Raymahashay†, outcrops of sulphides in the Darjeeling During the past thirty years, ground- Saumyen Guha†, Amitava Bhowmik† Himalayas, which contain up to 0.8% water has been used increasingly for irriga- *Geological Survey of , arsenic; and other sources in the upper tion and the use of phosphate fertilizers has Calcutta 700016, India reaches of the Ganges river system. increased threefold. More than 0.5 million †Department of Civil Engineering, The Ganges alluvial tract upstream of tubewells with handpumps, 0.1 million Indian Institute of Technology, Rajmahal, in the states of Bihar and Uttar shallow tubewells, and 3,000 deep tubewells Kanpur 208016, India

Pradesh, does not suffer from large-scale have been sunk at depths of 10–20 m, 1. Nickson, R. et al. Nature 395, 338 (1998). 8,9 arsenic contamination. This indicates that 30–100 m and 50–200 m, respectively . 2. Bhattacharya, P. et al. Int. J. Water Res. Dev. 13, 79–92 (1997). the Quaternary sediments around the This widespread withdrawal of ground- 3. Hiller, K. Geol. Jahrb. D 90, 3–35 (1988). Ganges delta have characteristic features water may have mobilized phosphate 4. Mukherjea, A. & Hazra, S. Ind. J. Geol. 69, 41–54 (1997). 5. Center for the Study of Man and the Environment, Calcutta that favour the initial retention and subse- derived from fertilizers and from the decay (unpublished data). quent release of arsenic. These sediments of natural organic materials in shallow 6. Central Groundwater Board, New Delhi (unpublished data). have high proportions of clay and contain aquifers. The increase in phosphate concen- 7. Geological Survey of India, Calcutta (unpublished data). relatively large amounts of organic carbon, tration could promote the growth of sedi- 8. Bagla, P. & Kaiser, J. Science 274, 174–175 (1996). 3,4 9. Mallick, S. & Rajagopal, N. R. Curr. Sci. 70, 956–958 (1996). and are thicker towards the south . The ment biota and the desorption of arsenic 10. Joshi, A. & Chaudhuri, M. J. Env. Eng. ASCE 122, 769–771 (1996). average arsenic content in cores from bore- from sediments. These combined microbio- 11.Wong, P. L. N., Huang, J. C. Y. & Cheng, T. W. J. Chin. Inst. Env. holes in West Bengal is higher in layers of logical and chemical processes might have Eng. 5, 241–251 (1995). clay (9.5–12 p.p.m.) than of sand (3.8–4.8 p.p.m.)5. Quaternary sediments in the Ganges alluvial tract in Bihar and Uttar Pradesh contain more sand and are much e have been studying the contami- Bangladesh groundwater. We disagree with narrower than sediments in the Bengal nation of groundwater by arsenic Nickson et al.’s claim that arsenic concen- basin, which may explain why they retain Wand the attend-ant human suffering trations in shallow (oxic) wells are mostly less arsenic. in West Bengal, India, for a decade, and in below 50 Ȗg per litre. In our samples from The groundwater of Uttar Pradesh and Bangladesh for the past four years. From Bangladesh (nǃ9,465), 59% of the 7,800 Bihar has trace concentrations of iron (0 our analysis of thousands of samples of samples taken at known depth and contain- to 0.7 mg per litre) compared with higher water and sediment1–7, we have been able to ing arsenic at over arsenic 50 Ȗg per litre values in West Bengal (up to 36 mg per test the course of events proposed by Nick- were collected from depths of less than 30 litre) and Bangladesh (30 mg per litre)1. son et al.8 to account for the poisoning of m, and 67% of the 167 samples with arsenic

NATURE | VOL 401 | 7 OCTOBER 1999 | www.nature.com © 1999 Macmillan Magazines Ltd 545 brief communications

10 6.4-9.7 in iron, as we and others have found in 11-15.8 Bangladesh.

) 20-21.9 Tarit Roy Chowdhury*, Gautam Kumar m ( 24.7-28

e Basu*, Badal Kumar Mandal*, Bhajan g

n 29.3-40.2 a Kumar Biswas*, Gautam Samanta*, Uttam r

h 38.4-40.2 t

p Kumar Chowdhury*, Chitta Ranjan

e 43-52.4 D n = 7,800 Chanda*, Dilip Lodh*, Sagar Lal Roy*, 56.7-89 Khitish Chandra Saha*, Sibtosh Roy†, 93.3-150 >150 Saiful Kabir†, Qazi Quamruzzaman†, 0 10 20 30 40 50 60 70 Dipankar Chakraborti* Samples (%) *School of Environmental Studies, Figure 1 The proportion of well samples containing more than Figure 2 Photomicrograph (magnification, ǂ100) showing the Jadavpur University, Calcutta 700032, India 50 Ȗg per litre arsenic changes with depth in Bangladesh. arsenic-containing opaque particles that are abundant in sedi- e-mail: [email protected] ments. Analysis of selected opaque particles by X-ray diffraction † Community Hospital Trust, concentrations above 1,000 Ȗg per litre and laser microprobe mass analysis identified the minerals pyrite, Bara Moghbazar, Dhaka 1217, Bangladesh rozenite (FeSO .4H O), haematite and magnetite in association were collected from wells between 11 and 4 2 1. Das, D. et al. Analyst 120, 917–924 (1995). 15.8 m deep. with quartz and calcite. 2. Das, D. Thesis, Jadavpur Univ., Calcutta, India (1995). In the district of Bangladesh, 3. Das, D. et al. Environ. Geochem. Health 18, 5–15 (1996). 4. Dhar, R. K. et al. Curr. Sci. 73, 48–59 (1997). three of the shallowest tubewells (depths of per kg. Our microscopic examination of 5. Samanta, G. et al. Microchem. J. 62, 174–191 (1999). 6.4 to 9.7 m) contained arsenic concentra- sediment samples from areas with arsenic- 6. Mandal, B. K. et al. Sci. Tot. Environ. 218, 185–201 (1998). tions of over 1,000 Ȗg per litre. The shal- contaminated groundwater indicates an 7. Roy Chowdhury, T. Thesis, Jadavpur Univ., Calcutta, India (1999). 8. Nickson, R. et al. Nature 395, 338 (1998). lowest tubewell in Bangladesh is 6.4 m, and abundance of these opaque particles (Fig. 9. Nickson, R. Thesis, University College London (1997). at that depth in the village of Chandipur, at 2). Electron microprobe analysis of the par- 10.Mott MacDonald Ltd British Geological Survey Final Report the Ramganj police station, we found an ticles revealed increased iron, sulphur and (January 1999). Ȗ 3,6 11.Akai, J. et al. in Conf. Proc. 3rd Forum Arsenic Contamination of arsenic concentration of 1,354 g per litre. arsenic , with arsenic content ranging from Groundwater in 51–54 (Faculty of Engineering, Miazaki 7 In the , we found arsenic at 0.07 to 1.36% by mass . University, Japan, 1998). 2,700 Ȗg per litre at a depth of 9.7 m. As Laser microprobe mass analysis identi- shown in Fig. 1, many of the shallowest fied arsenic-rich pyrite3, and X-ray diffrac- McArthur replies — In their comments on wells contain more than 50 Ȗg per litre of tion identified the minerals pyrite, rozenite our paper1, both Acharyya et al. and ǃ arsenic. In West Bengal, (n 55,000), we (FeSO4.4H2O), haematite and magnetite Chowdhury et al. refer to previous sugges- found arsenic above 50 Ȗg per litre in 28% in association with quartz and calcite3,7. tions2,3 that iron reduction might be the of wells that were less than 34 m deep. Fur- Pyrite was found7 in fine to coarse subangu- source of arsenic pollution. Naming a thermore, 37 of 38 samples with more than lar/subrounded sands and occasionally with process2, suggesting that it occurs on the 1,000 Ȗg per litre of arsenic were collected rock fragments at depths from 10 to 221 m. basis of reasoned argument3, and providing from depths shallower than 34 m. Arsenic has previously been found in objective evidence that it does1,4 are differ- Nickson et al.8 reported that arsenic con- pyrites from sediments of Bangladesh9,11. ent things. centration increases with depth in wells at But the question remains over whether Unpublished data from both sets of Manikganj, Faridpur and Tungipara, appar- arsenic-rich pyrite is sufficient to account authors confirm reports5–7 that concentra- ently on the basis of only a small number of for the mass of arsenic mobilized in tions of dissolved arsenic peak at depths of samples9. From studies of 320 tubewells in groundwater. It is often argued that if pyrite 20 to 40 m at some localities in the Ganges these areas, we find that arsenic concentra- is oxidized, the aquifer should be rich in plain. These maxima appear in data aggre- tion increases with depth for depths of less sulphate, but it is not. However, the forma- gated from separate wells dispersed across than 22 m and decreases at depths of over tion of framboidal pyrite, which has been unspecified areas, so their existence needs 22 m (Fig. 1), and we have observed a simi- identified9,11 in Bangladesh, might explain to be confirmed by vertical profiling at sin- lar trend in West Bengal. The British Geolog- the lack of sulphate in groundwater. gle localities. Nevertheless, these important ical Survey has recently reported that deep Nonetheless, we are not convinced that a data indicate that our findings that arsenic tubewells are free of arsenic in Bangladesh10. single mechanism is responsible for the concentration increases with depth1,4 at We agree with Nickson et al. that arsenic release of arsenic from aquifer sediments in some sites may not be a typical trend. associated with iron oxyhydroxides may be this , and believe that both the pyrite- It has been reported6 that 1% of wells leached from the sediments under reducing oxidation and iron oxyhydroxide-reduction deeper than 200 m are contaminated with conditions. However, we have suggested6 hypotheses deserve further consideration. arsenic. We reported1 that arsenic concen- that this could be one of several ways in We agree with Nickson et al. that trations were mostly below 50 Ȗg per litre which arsenic is leached from aquifer sedi- groundwater in Bangladesh with high iron in oxic (shallow) wells, not shallow (oxic) ments. We also proposed that the reducing concentrations can be passively ‘treated’ by wells, as stated by Chowdhury et al.: the dif- environment in the younger region of the allowing oxidation, precipitation and set- ference is that depth control on concentra- Ganges delta may have resulted in the for- tling of the ambient iron and the concomi- tions of dissolved arsenic is subordinate to mation there of arsenic-rich pyrite and that tant removal of arsenic. We have observed redox control. The statement6 that “shallow arsenic is mobilized by the oxidative disso- 60–70% arsenic removal in samples of well hand-dug wells are usually uncontaminated” lution of these solids3,6,7. water in which a brown precipitate (pre- is not surprising as they are likely to be This theory is based on our analysis of sumably iron oxyhydroxide) had formed oxic. As neither Acharyya et al. nor Chowd- 2,235 sediment samples, taken at intervals over time2. However, the efficiency of hury et al. have provided data on the redox of 3 or 6 m to a depth of 250 m, from 112 arsenic removal depends on the oxidation state of the water they analysed, their data boreholes in West Bengal. Of these samples, of arsenic and the concentration of iron, throw no light on these findings. 85 contained arsenic, ranging from 10 to and arsenic and iron occur together in some Pyrite oxidation2 is unlikely to be a 196 mg per kg. Opaque particles separated but not all samples. The proposed treat- major source of arsenic to anoxic water, from the parent sediment contained ment would therefore not be effective for which is the prevalent type beneath the arsenic3,6,7 at concentrations up to 2,778 mg samples that are high in arsenic but low southern Ganges plain. Oxidation of pyrite

546 © 1999 Macmillan Magazines Ltd NATURE | VOL 401 | 7 OCTOBER1999 | www.nature.com